# --- Core Libraries ---
import pandas as pd
import numpy as np
import networkx as nx
import os
import time
import glob

# --- Visualization Libraries ---
import matplotlib.pyplot as plt
import matplotlib.dates as mdates

# --- Notebook Settings ---
# Set a consistent style for all plots
plt.style.use('seaborn-v0_8-whitegrid')
# Make plots appear inline
%matplotlib inline
# Set a higher DPI for clearer plots
plt.rcParams['figure.dpi'] = 100
plt.rcParams['savefig.dpi'] = 300

print("Libraries imported successfully.")

Libraries imported successfully.

import os

# --- IMPORTANT: Set the project root directory ---
PROJECT_ROOT = r'D:\Users\Dsolv\Documents\Undergraduate\2025 spring\NSC\Proj' 

try:
    os.chdir(PROJECT_ROOT)
    print(f"Successfully changed working directory to: {os.getcwd()}")
except FileNotFoundError:
    print(f"Error: The directory '{PROJECT_ROOT}' was not found. Please update the path.")

# Define file paths
DATA_DIR = 'data'
REPORT_DIR = 'report'
FIGURES_DIR = os.path.join(REPORT_DIR, 'figures_notebook') # Create a new dir for notebook figures
os.makedirs(FIGURES_DIR, exist_ok=True)

MASTER_TRANSACTIONS_FILE = os.path.join(DATA_DIR, 'master_transfers.parquet')
CORRECTED_METRICS_FILE = os.path.join(REPORT_DIR, 'daily_network_metrics_corrected.csv')

# Load the data
try:
    master_df = pd.read_parquet(MASTER_TRANSACTIONS_FILE)
    metrics_df = pd.read_csv(CORRECTED_METRICS_FILE, index_col=0, parse_dates=True)
    
    # Combine USDC and USDT for 'flight to safety' analysis
    if 'volume_usdc' in metrics_df.columns and 'volume_usdt' in metrics_df.columns:
        metrics_df['volume_safe_stables'] = metrics_df['volume_usdc'] + metrics_df['volume_usdt']

    print("\nDaily Metrics Data Head:")
    display(metrics_df.head())
except FileNotFoundError as e:
    print(f"Error: Could not find a required data file.")
    print(f"-> {e}")

Successfully changed working directory to: D:\Users\Dsolv\Documents\Undergraduate\2025 spring\NSC\Proj
Successfully loaded pre-processed data.

Daily Metrics Data Head:

# This is the plotting helper function we will use for all time-series charts.

def plot_time_series(df, columns, title, ylabel, filename, use_log_scale=False, y_formatter=None):
    """
    Generic function to plot time series data, with options for log scale and custom y-axis formatting.
    """
    fig, ax = plt.subplots(figsize=(14, 7))
    
    for col in columns:
        # Create a more readable label for the legend
        label = col.replace('_', ' ').replace('volume', '').upper().strip()
        ax.plot(df.index, df[col], label=label, alpha=0.9)
    
    # Use logarithmic scale if specified
    if use_log_scale:
        ax.set_yscale('log')
        ylabel += " (Log Scale)"
    
    # Apply a custom formatter for the y-axis if provided
    if y_formatter:
        ax.get_yaxis().set_major_formatter(plt.FuncFormatter(y_formatter))
    
    # Formatting the date axis
    ax.xaxis.set_major_locator(mdates.MonthLocator(interval=1))
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %Y'))
    
    # Adding a vertical span to highlight the main crisis period
    crisis_start = pd.to_datetime('2022-05-08')
    crisis_end = pd.to_datetime('2022-05-20')
    ax.axvspan(crisis_start, crisis_end, color='red', alpha=0.15, label='Crisis Period (May 8-20)')

    # Adding a vertical line for the FTX collapse for context
    ftx_collapse = pd.to_datetime('2022-11-08')
    ax.axvline(ftx_collapse, color='black', linestyle='--', linewidth=1.2, alpha=0.7, label='FTX Collapse')

    ax.set_title(title, fontsize=18, pad=15)
    ax.set_ylabel(ylabel, fontsize=12)
    ax.set_xlabel("Date", fontsize=12)
    ax.legend()
    fig.autofmt_xdate()
    ax.grid(True, which="both", ls="--", alpha=0.5)
    
    # Save the figure to our notebook's dedicated figure directory
    plt.savefig(os.path.join(FIGURES_DIR, filename), dpi=300, bbox_inches='tight')
    plt.show()

# A formatter to make large numbers on the y-axis more readable (e.g., 100,000, 1B)
def human_readable_formatter(x, pos):
    if x >= 1e9:
        return f'{x*1e-9:1.0f}B'
    if x >= 1e6:
        return f'{x*1e-6:1.0f}M'
    if x >= 1e3:
        return f'{x*1e-3:1.0f}K'
    return f'{x:1.0f}'

# Plot 1: Daily Network Activity
plot_time_series(metrics_df, 
                 columns=['nodes', 'edges'], 
                 title='Daily Network Activity: A Surge During Crisis', 
                 ylabel='Count',
                 filename='plot_01_daily_activity.png',
                 y_formatter=human_readable_formatter)

# Plot 2: Daily Transaction Volume of Terra Ecosystem Tokens
plot_time_series(metrics_df, 
                 columns=['volume_wluna', 'volume_ustc'], 
                 title='The Death Spiral: Terra Ecosystem Transaction Volume (USD)', 
                 ylabel='Transaction Volume (USD)',
                 filename='plot_02_terra_volume.png',
                 use_log_scale=True,  # Log scale is essential here
                 y_formatter=human_readable_formatter)

# Plot 3: Daily Transaction Volume of "Safe" Stablecoins
plot_time_series(metrics_df, 
                 columns=['volume_safe_stables'], 
                 title='The Flight to Safety: Volume of USDC & USDT', 
                 ylabel='Transaction Volume (USD)',
                 filename='plot_03_safe_stables_volume.png',
                 y_formatter=human_readable_formatter)

# Plot 4: Daily Average Clustering Coefficient
plot_time_series(metrics_df, 
                 columns=['avg_clustering'], 
                 title='Network Structure: Daily Average Clustering Coefficient', 
                 ylabel='Clustering Coefficient',
                 filename='plot_04_avg_clustering.png')

# --- Setup for Deep Dive Analysis ---
DATE_NORMAL = '2022-05-04'
DATE_PANIC = '2022-05-09'

# The key exchange wallet we want to investigate (Binance 8 Wallet)
HUB_TO_ANALYZE = '0x56178a0d5f301baf6cf3e1cd53d9863437345bf9'

# Load the necessary price data for on-the-fly correction
PRICE_DF_WLUNA = pd.read_csv(os.path.join(DATA_DIR, 'price_data', 'wluna_price_data.csv'))
PRICE_DF_WLUNA['date'] = pd.to_datetime(PRICE_DF_WLUNA['timestamp'], unit='s').dt.date

def analyze_hub_net_flow_corrected(hub_address, date_str):
    """
    Calculates the USD net flow for a hub on a specific day by
    querying the master transaction dataframe and applying price correction for WLUNA.
    """
    # Filter master DF for the specific day
    target_date = pd.to_datetime(date_str).date()
    day_df = master_df[pd.to_datetime(master_df['time_stamp']).dt.date == target_date]
    
    # Get the WLUNA price for that specific day
    try:
        day_price_wluna = PRICE_DF_WLUNA[PRICE_DF_WLUNA['date'] == target_date]['close'].iloc[0]
    except IndexError:
        day_price_wluna = 0

    net_flow = {'WLUNA': 0, 'USTC': 0, 'USDC': 0, 'USDT': 0}
    
    # Filter for transactions involving the hub
    in_df = day_df[day_df['to_address'] == hub_address]
    out_df = day_df[day_df['from_address'] == hub_address]
    
    for token in net_flow:
        in_value = in_df[in_df['token_name'] == token]['value'].sum()
        out_value = out_df[out_df['token_name'] == token]['value'].sum()
        
        # We only need to price-correct WLUNA here as stablecoins are ~$1
        if token == 'WLUNA':
            in_value *= day_price_wluna
            out_value *= day_price_wluna
        
        net_flow[token] = in_value - out_value
        
    return net_flow

print("Deep dive analysis functions are ready.")

Deep dive analysis functions are ready.

# --- Execute the Net Flow Analysis ---
print(f"Analyzing Net Flow for Key CEX Wallet: {HUB_TO_ANALYZE}")

flow_normal = analyze_hub_net_flow_corrected(HUB_TO_ANALYZE, DATE_NORMAL)
flow_panic = analyze_hub_net_flow_corrected(HUB_TO_ANALYZE, DATE_PANIC)

df_flow = pd.DataFrame([flow_normal, flow_panic], index=[DATE_NORMAL, DATE_PANIC])

print("\n--- Net Flow Table (Positive = Net In-flow) ---")
# Use a custom formatter for the table display
display(df_flow.style.format("{:,.0f}"))


# --- Plotting the Net Flow Bar Chart ---
fig, ax = plt.subplots(figsize=(12, 7))
df_flow.plot(kind='bar', ax=ax, rot=0, width=0.8)

ax.set_title('The Exit Ramp: Token Net Flow for Binance 8 Wallet', fontsize=18, pad=15)
ax.set_ylabel('Net Volume (USD)')
ax.set_xlabel('Date')
ax.axhline(0, color='black', linewidth=0.8)
ax.get_yaxis().set_major_formatter(plt.FuncFormatter(human_readable_formatter))
ax.grid(axis='y', linestyle='--', alpha=0.7)

# Make legend labels more readable
handles, labels = ax.get_legend_handles_labels()
ax.legend(handles, [l.upper() for l in labels], title='Token')

plt.tight_layout()
plt.savefig(os.path.join(FIGURES_DIR, 'plot_05_binance_net_flow.png'))
plt.show()

Analyzing Net Flow for Key CEX Wallet: 0x56178a0d5f301baf6cf3e1cd53d9863437345bf9

--- Net Flow Table (Positive = Net In-flow) ---

# --- Degree Distribution Comparison ---
print("Loading graphs for degree distribution analysis...")

# Define the graph loading and plotting functions
def get_total_value_graph(G_rich):
    """Creates a simplified graph with a single 'weight' attribute for total value."""
    G_simple = nx.DiGraph()
    for u, v, data in G_rich.edges(data=True):
        total_weight = sum(w for w in data.values() if isinstance(w, (int, float)))
        if total_weight > 0:
            G_simple.add_edge(u, v, weight=total_weight)
    return G_simple

def plot_degree_distribution(G, date_str, ax):
    """Plots the log-log degree distribution on a given axis."""
    degrees = [d for n, d in G.degree() if d > 0]
    if not degrees: return
    
    max_deg = max(degrees)
    if max_deg < 2: return
    # Use more bins for a clearer plot
    bins = np.logspace(np.log10(1), np.log10(max_deg), 50)
    
    hist, bin_edges = np.histogram(degrees, bins=bins, density=True)
    bin_centers = (bin_edges[:-1] + bin_edges[1:]) / 2
    
    non_zero = hist > 0
    ax.loglog(bin_centers[non_zero], hist[non_zero], 'o', alpha=0.7, markersize=5, label=f'Degree Dist. ({date_str})')

# Load the full daily graphs
G_normal = nx.read_gexf(os.path.join('data', 'processed', 'daily_graphs', f"graph_{DATE_NORMAL}.gexf"))
G_panic = nx.read_gexf(os.path.join('data', 'processed', 'daily_graphs', f"graph_{DATE_PANIC}.gexf"))

# Create the plot
fig, ax = plt.subplots(figsize=(12, 7))
plot_degree_distribution(G_normal, DATE_NORMAL, ax)
plot_degree_distribution(G_panic, DATE_PANIC, ax)

ax.set_title("Degree Distribution Shifts During Panic", fontsize=18, pad=15)
ax.set_xlabel("Degree (k) - Number of Connections")
ax.set_ylabel("Probability P(k)")
ax.grid(True, which="both", ls="--", alpha=0.5)
ax.legend()

plt.tight_layout()
plt.savefig(os.path.join(FIGURES_DIR, 'plot_06_degree_distribution.png'))
plt.show()

Loading graphs for degree distribution analysis...

	nodes	edges	avg_clustering	volume_dai	volume_pax	volume_usdc	volume_usdt	volume_ustc	volume_wluna	volume_safe_stables
2022-04-01	56913	75007	0.010599	5.279960e+08	6.818257e+07	5.998725e+09	3.112786e+09	3.633266e+08	2.010514e+05	9.111511e+09
2022-04-02	80010	109034	0.012547	1.245802e+09	1.138278e+07	4.982127e+09	3.006956e+09	2.579704e+08	2.679326e+07	7.989082e+09
2022-04-03	70555	95065	0.010539	7.853623e+08	7.336793e+06	3.252445e+09	1.998822e+09	4.889970e+09	4.046141e+07	5.251266e+09
2022-04-04	87631	120695	0.014743	6.531790e+08	2.601105e+07	1.026462e+10	3.936265e+09	3.431633e+08	3.464580e+07	1.420089e+10
2022-04-05	82777	113852	0.011573	7.588290e+08	1.162615e+08	8.797779e+09	4.251814e+09	2.275106e+08	4.613037e+07	1.304959e+10

Network Analysis of the Terra/LUNA Collapse¶

Introduction¶

Part 0: Data Ingestion and Preprocessing¶

0.1. Justification for Data Unification¶

0.2. Price Correction and USD Valuation¶

0.3. Loading the Processed Data¶

Part I & II: Time-Series Analysis - The Network's Vital Signs¶

Analysis of Network Activity¶

Analysis of the Terra Ecosystem's Collapse¶

Analysis of the "Flight to Safety"¶

Analysis of Network Structural Cohesion¶

Part III: Deep Dive - Anatomy of a Panic¶

Analysis of the "Exit Ramp": The Binance 8 Wallet¶

Analysis of Structural Change: Degree Distribution¶

Network on a Normal Day: May 4, 2022¶

Network on a Panic Day: May 9, 2022¶

Conclusion: A Network Perspective on Systemic Risk¶

	WLUNA	USTC	USDC	USDT
2022-05-04	43,021	0	-2,016,693	-4,861,342
2022-05-09	-20,819	0	2,845,062	8,905,253